Year: 2025
Type: article
Abstract: Generative Artificial Intelligence (GenAI) is rapidly transforming the software testing landscape, introducing both groundbreaking opportunities and significant challenges. Traditional testing techniques are increasingly inadequate for evaluating GenAI systems, which generate novel, diverse, and often unpredictable outputs. This has led to fundamental issues such as the "oracle problem," difficulties in test adequacy assessment, and concerns about bias, privacy, and explainability. Across academic and industry perspectives, researchers and practitioners highlight the need for new methodologies—such as metamorphic testing, differential testing, and diversity- based adequacy measures—to address the unique complexities of GenAI. Moreover, the role of AI in supporting test generation, prioritization, and automation is expanding, promising increased efficiency and scalability. However, ensuring trustworthiness, accountability, and ethical deployment of GenAI in critical domains like healthcare and finance requires careful integration of human oversight, rigorous validation techniques, and the development of interpretable models. This body of work collectively underscores the urgent need for interdisciplinary efforts to develop robust, adaptive, and transparent testing frameworks tailored for GenAI systems.
Authors: R. Lavanya Bai, C. H. Saradadevi, S. Shri Preetha
Institution: Meenakshi Academy of Higher Education and Research
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4410867214